Skip to main content

A new method for cross-cultural and cross-temporal comparison of societies

Final Report Summary - IMGAME (A new method for cross-cultural and cross-temporal comparison of societies)

We developed a new method of research in the social sciences and discovered some surprising results. The new method is called the Imitation Game. We had been exploring this method since before the turn of the century but on a small scale; the project developed its use on a large scale.

The imitation game is what gave Alan Turing, the computer pioneer, the idea for his famous Turing Test for computer intelligence: a ‘judge’ asks questions of a hidden person and a hidden computer and if it is not possible to distinguish between them we say the computer is ‘intelligent’. The game, which came first, was played between men and women, one pretending to be the other with the judge trying to work out who was who just from questioning while, say, the man pretended to be a woman. We had used the Imitation Game (we capitalize when we refer to one of our carefully worked out protocols), to show that, as we expected, the blind were better at pretending to be sighted than the sighted were at pretending to be blind; the reason we expected this was because the blind spend their life in conversation with the sighted but not vice versa, so the blind possessed what we call ‘interactional expertise’ in the life of the sighted. We had also used it as a test for individual researcher’s understanding of a respondent community – Collins could pass as a gravitational wave physicist, a community he was studying. The grant enabled us to extend this research to whole countries – we wanted to know, for example, whether secular people were better at pretending to be religious in, say, Sicily, where Catholicism is ‘in the air’ than in, say, Holland, where Christianity is much less socially ubiquitous – they were. Our idea was that this approach could help us understand the level of understanding of minorities in different countries and, perhaps, the level of understanding of men for women and vice versa in countries with different cultural ambiences. This required large samples and much organization and travel.

It also required an immense effort of computerization to control the large-scale experiments – much more than we anticipated – and many statistical and mathematical problems had to be solved: instead of three people in a room we had 24 people playing the game simultaneously taking the role of judge in one game, ‘pretender’ in another game and non-pretender in a third game. We also deconstructed the game so that the judge in the first round of games was really used only to generate questions while a much larger number of pretenders answered these questions in a second step. Then dialogues with these answers were reassembled with the non-pretenders answers and presented to real judges at the final step. This gave us much better statistical significance. We decided, however, that since this was a new method that needed to be proved we would only count results as sound if they were repeatable at least once and this meant the generation and confirmation of a finding was a drawn out process.

Not everything worked out as we expected. Quantitative data was harder to interpret than we expected and the method has to be used carefully with large samples as the Game is more complicated than it appears and the data contains a lot of noise. That said, we did find a number of interesting and unexpected results. For example, we found a higher tolerance for minority cultures in places we expected to be intolerant, only finding out later that our expectations were at fault and that the data were more robust than we initially thought. We also found that in South Africa whites understood black culture better than blacks understood white culture whereas we expected the opposite basing our ideas on what had been written about the relationship between the cultures in America. Perhaps most significantly of all, we discovered that the qualitative data generated alongside the quantitative data was much more interesting than we had first anticipated, with the content of questions and answers offering a different kind of insight into a culture than that provided by other research methods.